Kafka MCP Server bridges the gap between LLM models and Apache Kafka, enabling them to produce and consume messages, list, describe, and manage topics, monitor consumer groups, and assess cluster health and configuration through the MCP interface. The server implements an MCP Protocol Handler with three registries—Tools (Kafka operations), Resources (health and diagnostics), and Prompts (pre-configured workflows)—and a Kafka client wrapper that communicates with the cluster via the franz-go library. The architecture connects MCP clients (AI applications such as Claude Desktop, Cursor, Windsurf, and ChatWise) to an Apache Kafka cluster, translating MCP requests into Kafka API calls and returning results. It supports STDIO by default for local use and HTTP for remote access, with optional OAuth 2.1 authentication on the HTTP transport. The server exposes tools (produce_message, consume_messages, list_brokers, describe_topic, list_consumer_groups, describe_consumer_group, describe_configs, cluster_overview, list_topics), resources (overview, health_check, under-replicated_partitions, consumer_lag_report), and prompts (kafka_cluster_overview, kafka_health_check, kafka_under_replicated_partitions, kafka_consumer_lag_report). Security includes SASL, TLS, and OAuth 2.1 providers.